The reform that moved postings, not hires.

Skills-based hiring moved job postings, not jobs. The reform has a spectacular announcement record: degree requirements struck from postings by the thousand, coalition pledges signed, careers pages rewritten to promise that ability counts for more than diplomas. The posting changes were real. But when researchers at the Burning Glass Institute matched a decade of postings at large firms to the hires that actually followed, the celebrated reform shrank to "not even 1 in 700 hires," in the report's own phrasing (Sigelman, Fuller, & Martin, 2024). The announcements filled years of keynotes. The hiring ledger barely moved.

It would be easy to read that gap as hypocrisy, and some of it surely is. The more useful reading is mechanical. A degree requirement is a filter: crude, cheap, and wired into every screening workflow that touches an application. Deleting it removes a barrier without supplying a measurement, and a hiring funnel does not tolerate a vacuum. Screeners still have to decide, and deprived of the degree line they reach for the nearest available proxies, which behave much like the credential they replaced. The reform, as most firms practiced it, deleted a signal and installed nothing.

The evidence assembled below supports a sharper conclusion than either the celebration or the backlash. The problem skills-based hiring set out to fix was real and expensive. The posting reset was real and large. The hiring effect was real and tiny. And the minority of firms that turned the pledge into practice captured gains worth having: workers hired without degrees into formerly degree-gated roles earn about 25% more than they did before, over $12,400 a year, and stay longer than their degreed peers (Sigelman, Fuller, & Martin, 2024). What separates the announcement from the outcome is not sincerity but whether anyone built the instrument that was supposed to replace the deleted filter.

The degree filter was expensive, and employers knew it

Begin with why the reform attracted so much conviction, because its premise was never the weak point. In 2017, Harvard Business School's Joseph Fuller and Manjari Raman published Dismissed by Degrees, a study conducted with Accenture and Grads of Life that gave the underlying pathology a name: degree inflation, the practice of demanding a four-year degree for work that the people currently doing it perform without one (Fuller & Raman, 2017). The diagnosis rested on an analysis of 26 million job postings, read against the actual education of the employed workforce, occupation by occupation.

Figure 1 puts the report's signature comparison side by side. In 2015, 67% of postings for supervisors of production workers asked for a college degree, while 16% of the people employed in that occupation held one (Fuller & Raman, 2017). The posting described a workforce that did not exist. Whatever the degree line was screening for, it was not a description of the incumbents already doing the job to their employers' evident satisfaction.

The study sized the exposure at 6.2 million middle-skills jobs at risk of degree inflation, jobs demanding more preparation than high school but not historically a bachelor's degree. And it priced the habit. In a survey of 600 business and HR leaders, a majority reported paying 11–30% more for college graduates in roles also filled by non-degree holders, without material improvement in productivity (Fuller & Raman, 2017). Degree requirements were not a free conservatism; employers were paying a premium for a credential their own experience told them the work did not require.

The case landed. Within a few years, dropping degree requirements had become the rare talent policy endorsed by governors, chief executives, and workforce advocates at once, and skills-first hiring entered the standard vocabulary of corporate citizenship reports. The question was never whether the argument was persuasive. It was what, operationally, a firm does after it agrees.

Degree requirements did fall, exactly where everyone could see them

The first wave of evidence looked like confirmation. The Emerging Degree Reset, a 2022 Burning Glass Institute analysis of posting data, found that between 2017 and 2019, 46% of middle-skill occupations and 31% of high-skill occupations experienced material degree resets, the report's term for a substantial drop in the share of postings specifying a four-year degree (Fuller et al., 2022). Projecting the trend forward, the authors estimated 1.4 million additional jobs opening to workers without degrees over the following five years.

The reset was most visible at the top of the market. The share of Apple postings specifying a bachelor's degree or above fell from 88% to 72%, and Google's from 93% to 77% (Fuller et al., 2022). Accenture moved harder in the occupations it targeted: its postings for computer support specialists that specified a degree fell from 46% in 2017 to 9% in 2021. These are the employers whose practices get imitated, and the numbers traveled fast into keynotes, pledge campaigns, and board decks.

It is worth pausing on why this layer of the reform moved so quickly, because the speed itself is diagnostic. A posting is the cheapest component of a hiring system to change: editing a requirements line needs no budget, no new workflow, no training, and no one's approval beyond recruiting operations. Replacing what the requirement did (that is, standing up an assessment process that can tell a capable non-graduate from an incapable one) touches procurement, recruiter habit, hiring-manager authority, and time-to-fill targets. When an initiative's visible layer is nearly free and its operative layer is expensive, adoption statistics for the visible layer will always run ahead of the practice, and posting data will flatter every participant.

The same report carried a counterweight the keynotes tended to omit: 37% of middle-skilled jobs showed no reduction in degree requirements at all, keeping an estimated 15.7 million workers out of candidate pools (Fuller et al., 2022). Even at its posting-level peak, the reset was partial. Still, by any ordinary standard of institutional change this was movement, measured in language every applicant could read. What the posting data could not answer was the question underneath it: had anyone's hiring changed?

Matched to actual hires, the reform nearly disappears

That is the question the Burning Glass Institute's 2024 follow-up, Skills-Based Hiring: The Long Road From Pronouncements to Practice, was designed to answer (Sigelman, Fuller, & Martin, 2024). Instead of reading postings, the researchers matched them to career-history data for a sample of 11,300 roles at large firms from 2014 to 2023, so that every dropped requirement could be checked against the education of the people subsequently hired into the role. It is the difference between auditing the press release and auditing the payroll.

The design matters because the two data sources answer different questions. Posting studies count promises; career-history matching counts outcomes; and that space is where an initiative can be celebrated for years without ever being tested. And because the window opens in 2014, before the pledge era, changes in hiring can be read against how the same roles hired before the announcements began.

The within-role finding vindicates the reform's logic. Where a role genuinely dropped its degree requirement, the share of new hires without degrees rose: measurably, if modestly (Sigelman, Fuller, & Martin, 2024). Removing the line really does change who gets hired. The collapse happens one step upstream, in how few lines were ever struck: over the full decade, only about 3.6% of roles dropped a degree requirement at all, and a within-role gain applied to so marginal a slice of the economy nets out to a change of about 0.14 percentage points in the national hire mix. Figure 2 traces the shrinkage stage by stage, ending at a scale bar in which a full year of the reform's incremental hires nearly disappears into the annual total.

Averages this small can hide a split, and this one does. When the report classified the firms that had removed degree requirements by what their hiring subsequently did, 37% qualified as Skills-Based Hiring Leaders: firms whose hiring of non-degreed workers into the affected roles rose by about 18% (Sigelman, Fuller, & Martin, 2024). The largest group, 45%, was classified In Name Only, companies whose postings changed while their hiring did not. Figure 3 shows the full segmentation; note that these are shares of firms, not of postings or hires, so a single large In Name Only employer can outweigh several small Leaders in the aggregate numbers.

Read together, the figures relocate the story. The interesting divide is not between firms that believe in skills and firms that do not; it is between firms where the posting edit was connected to anything downstream and firms where it was the downstream. For a substantial minority, dropping the requirement was one visible move in an operational change that reached screening, assessment, and hiring. For the plurality, the posting was the entire intervention, and the hiring engine underneath it kept executing the old logic on the old signals.

Deleting a filter does not supply a signal

A firm that removes a requirement and then hires as if it had not is not necessarily insincere. It is running its own workflow along the path of least resistance. However weak as a predictor, the degree line answered a real operational question: with hundreds of applications per opening and minutes per screen, what do we use to say no? Selection research priced the credential proxies decades ago, with years of education correlating .10 with later job performance and years of experience .18 (Schmidt & Hunter, 1998), a record our companion analysis of résumé screening accuracy examines line by line. But in the daily practice of a screener with a queue, a weak signal that exists beats a strong signal that does not. Strike the degree line without installing a measurement, and the decision migrates to whatever remains on the page: employer brand names, title sequences, years logged. Pedigree by other means.

The filter also survives in places a posting edit never reaches. Requisition templates carry the degree field forward from the last opening; applicant tracking systems keep the knockout question that was configured years earlier; hiring managers retain a mental model of what a plausible candidate looks like, formed on a decade of degreed shortlists. A recruiter judged on time-to-fill has no incentive to champion an unfamiliar profile a manager may bounce. Each of these is a small, reasonable-seeming default, and together they reconstitute the deleted requirement informally, one stage downstream of wherever the reform stopped.

The posting data itself records the vacuum. The 2022 reset analysis found that when employers drop degree requirements, their postings become more specific about skills, spelling out demands for writing, communication, and detail orientation that the degree had been assumed to certify (Fuller et al., 2022). That is the demand for a replacement signal, visible in the language of the ask. Employers evidently know what the credential was standing in for. What most had not done, on the 2024 evidence, was adopt any instrument that measures those attributes rather than describing them in a longer paragraph.

The gap has an audience, which is what makes it a risk rather than a rounding error. Candidates who answered a skills-first posting and lost to pedigree can read the pattern in their own rejections. Boards that approved the pledge can read it in workforce composition that refuses to move. And any regulator or journalist with posting data and hiring data can read it the way the Burning Glass Institute did, by putting the two side by side. An announced reform that hiring records contradict is worse than no announcement, because it converts a policy question into a credibility question.

Skills-based hiring is a measurement project or it is a press release

What did the Leaders build that the In Name Only firms did not? The matched data does not follow individual firms into their assessment stacks, but the logic of the gap points in one direction. If the degree was a proxy for a bundle of unmeasured attributes, then making its removal real requires measuring those attributes. Treated seriously, skills-based hiring is a measurement project: for every requirement deleted, the organization specifies what the requirement was standing in for and adopts an instrument that assesses that thing directly, on the same terms, for every applicant.

The instruments are not exotic, and their evidence base is mature. Under the 2022 re-estimates by Sackett and colleagues, which applied less aggressive range-restriction corrections than earlier syntheses, job knowledge tests predict job performance at .40 and work samples at .33 (Sackett et al., 2022), each far above the credential proxies they would replace. The full ranking, and what it implies for assembling a selection process, is the subject of our review of what predicts job performance.

In practice, verification runs as a gradient of increasingly direct evidence. A claim on an application is the floor: unpriced, unverified, free to assert. Standardized testing of ability and job knowledge is the first genuine measurement, cheap enough to apply to a whole funnel. Standardization is also what makes the evidence comparable: every applicant faces the same tasks under the same conditions, scored the same way, so a candidate who learned the work outside a classroom produces the same kind of record as one who learned it inside. And the escalation only informs a decision if each stage leaves a score, not an impression. A work sample or simulation goes further and has the candidate perform a slice of the actual job. At the high end, the multi-exercise observation used in assessment centers watches candidates work extended, role-realistic problems under trained evaluation. How far up the gradient a role justifies is a cost question that scales with the stakes of the seat. That there must be some next step after the deleted requirement is the practical lesson of the whole literature: a funnel needs a basis for decisions, and the only real choice is between a measured basis and an improvised one.

The pool this unlocks is not a niche. Opportunity@Work counts more than 70 million U.S. workers as STARs, Skilled Through Alternative Routes: people with a high school diploma or equivalent, no four-year degree, and skills for higher-wage work (Opportunity@Work & Accenture, 2020). A default degree screen excludes the entire population before any evidence is taken. Hiring without a degree requirement, run on real instruments, is how members of that pool reach an interview slate at all, and it is the only version of the reform they can distinguish from the announcement.

For the employers that follow through, the 2024 report's payoff numbers are the strongest part of the story. Non-degree workers hired into formerly degree-gated roles earned an average salary increase of about 25%, over $12,400 a year, a measure of how much latent capability the old filter had been leaving in lower-paid work. Their employers kept them, too: two-year retention ran 10 percentage points higher than degreed coworkers in the same roles, 58% against 48% (Sigelman, Fuller, & Martin, 2024). A hire that represents a step up is repaid in tenure, and tenure is where hiring economics are won and lost.

Report the hires, not the postings

For an organization that has announced skills-first hiring, or is about to, the evidence reduces to a short operating agenda. Each item exists because a specific failure in the record above shows what happens without it.

1. Audit the gap between your postings and your payroll. List the roles where degree requirements were dropped, then pull the education mix of the people hired into those roles since. This is the same comparison the 2024 matched analysis ran at scale, and it is available to any employer from its own records. If the posting changed and the hire mix did not, the reform so far has been typographical, and it is better to learn that internally than from an outside analyst.
2. Name the replacement signal for every deleted filter. A dropped requirement should have a written successor: the specific instrument (a job knowledge test, a work sample, a structured behavioral assessment) that now measures what the degree was assumed to certify for that role. Administering standardized measurement at funnel scale is precisely what an assessment platform exists to make routine, and a disciplined way to evaluate candidate instruments is laid out in our guide to comparing assessment providers.
3. Instrument the funnel so the claim is testable. Track candidates without degrees through each stage and find where they exit. If they clear the assessment and vanish at the recruiter screen or the hiring-manager review, the old filter has reconstituted itself informally, and the fix is a scored rubric at that stage, not another posting edit.
4. Report the hires. The posting count is an input metric; the board, the workforce, and increasingly the public will judge the program on outcomes. Reporting the non-degree hire share for reformed roles, on a cadence, makes the commitment self-auditing and denies the gap the years of quiet accumulation that produced the 2024 headline.

The reform's first decade produced an unflattering natural experiment: thousands of firms changed their language, a minority changed their measurement, and only the second group changed their workforce. Read as evidence about the idea itself, the experiment is favorable: where the pledge was operationalized, hiring moved, pay rose, and retention improved. The organizations that treat a deleted requirement as a measurement vacancy, and fill it, are the ones whose reform shows up where it counts: not in the announcement economy, but in the hiring ledger.